Automatizirana provjera sukladnosti politika privatnosti na webu s Općom uredbom o zaštiti podataka korištenjem strojnog učenja

master's thesis

• Title (english): Automated compliance check of website privacy policies with the general data protection regulation using machine learning
• Author: Rene Šeša
• Mentor: Marko Horvat (mentor)
• Committee member: Marko Horvat (predsjednik povjerenstva)
• Committee member: Gordan Gledec (član povjerenstva)
• Committee member: Boris Vrdoljak (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-09, Croatia

Abstract

U ovome radu istražuju se mogućnosti automatiziranja procesa provjere sukladnost politika privatnosti s Općom uredbom o zaštiti podataka korištenjem metoda strojnog učenja. Odabrano je 5 odredbi GDPR koje su pogodnih za analizu koje svaka politika privatnosti mora imati: kontakt voditelja obrade, pravo ispitanika na povlačenje privole, navedenu svrhu obrade podataka, navedenu kategoriju osobnih podataka koji se obrađuju te obavijest ispitaniku na pravo na brisanje i ispravak prikupljenih podataka. Prilikom prikupljanja politika privatnosti korištena je Python biblioteka Polipy. Za vektorsku reprezentaciju riječi korišteni su modeli OpenAI text-embedding-3-small i SPLADE++. Navedeni modeli su trenirani za engleski jezik pa je za prijevod teksta s hrvatskog jezika na engleski jezik korištena biblioteka Translators. Za klasifikaciju paragrafa politika privatnosti korištene su metode strojnog učenja: logistička regresija i linearna regresija. Na kraju je donesen zaključak na temelju dobivenih rezultati te su navedena daljnja razmatranja.

Abstract (english)

This paper explores the possibility of automating the process of checking privacy policy compliance with the General Data Protection Regulation by using machine learning methods. 5 parts of the GDPR were selected that are suitable for analysis and that every privacy policy needs to have: contact of the data controller, notification of the subject's right to withdraw the consent, specification of the purpose of data processing, specification of the category of personal data that is processed, and notification to the data subject of the right to delete and correct collected data. The Python library Polipy is used to collect privacy policies. OpenAI text-embedding-3-small and SPLADE++ models are used for text embedding. The mentioned models are trained for English language usage, so the Translators library is used to translate the text from Croatian to English. Machine learning methods are used to classify privacy policy paragraphs: logistic regression and linear regression. Ultimately, a conclusion was made based on the obtained results, and further considerations were given.

• Keywords: GDPR; sukladnost; obrada prirodnog jezika; vektorska reprezentacija riječi; automatizacija
• Keywords (english): GDPR; compliance; natural language processing; text embedding; automation
• Language: croatian
• URN:NBN: urn:nbn:hr:168:679581
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 22:15:05